AI Agency · Colombia
Artificial intelligence agency in Colombia
An engineering team that integrates AI with the systems you already operate, not a pilot that stays a demo.
Coresis is a software engineering company in Colombia that works with artificial intelligence as one more capability in the stack. The difference from a recent AI agency is the starting point: we arrive from integration with production systems — institutional portals, intranets, platforms with sensitive data and traceability requirements — not from building prototypes.
That shapes how we design: before choosing a model, we review what process needs to be solved, which systems need to be touched, what data can leave the perimeter, and how the result will be measured.
An engineering team
with mission-critical platforms
in production
The credibility of an AI agency in Colombia shouldn't be measured by how many tools it mentions, but by whether it has sustained real systems over time. Coresis has implemented and sustained platforms for the Ministry of Housing, ICBF, USPEC, ICDE, Metro de Bogotá, and Corporación Colombia Digital, among other organizations.
That track record is what provides the judgment an AI project needs: integration with legacy systems, change control, staging environments, and a realistic idea of what it costs to operate something after launch. You can review the details in our portfolio.
What we build
- Integrated AI agents: assistants that query and act on CRM, ERP, databases, and our own APIs. The full technical explanation is in artificial intelligence agents.
- Process automation: flows that replace repetitive manual tasks, with rules where rules are enough and a language model only where it adds value.
- Knowledge bases with RAG: answers backed by the client's own documentation, versioned and updatable by the client's team.
- Knowledge graphs: when the domain has entities and relationships a similarity search doesn't capture well, we can model it as a graph so the agent reasons about relationships, not just text fragments.
- Integrations and interoperability: connecting to existing systems, which is usually the part that determines the project's real timeline.
The data constraint
shapes the design
In banking, healthcare, and the public sector, the question that shapes the project isn't which model to use, but what data can leave the infrastructure. That's why we work with orchestration in n8n, which can be self-hosted, and with open models deployed on our own infrastructure when sensitivity or volume justifies it. To build and version agents and knowledge bases we use Dify, as in Labbu. When delegated work requires retries, recovery from partial failures, and evidence of every step, we run it on AquilaMesh, our own runtime. If the case calls for a multi-agent system with no platform in between, we work with CrewAI or custom builds on LangChain. The choice depends on the case, not on preference.
When commercial models are used via API, that guarantee doesn't apply, and it's something worth deciding explicitly at the start rather than discovering at the end.
How we choose the stack:
the license is also
a technical decision
Many of the AI tools marketed as "open source" aren't, strictly speaking. Their code is available, but with restrictions that matter right when the project grows. It's worth reviewing this before building, not after.
- Dify is distributed under a modified version of Apache 2.0 that doesn't allow operating a multi-tenant environment without written authorization, or removing its branding from the console. For an internal product this isn't a problem; for a SaaS serving several organizations, it is.
- n8n uses the Sustainable Use License: it allows internal use, but not offering it as a commercial service to third parties. Its enterprise features require a separate license.
- CrewAI and LangChain are MIT-licensed, with no such restrictions.
- Commercial models via API also impose their own terms on data and usage.
This isn't an argument against any tool — we use several of them. It's that the license shapes the client's business model, and that conversation is worth having when choosing the stack, not once the product is already in production.
Projects with AI components
Coresis builds with AI in its own products before proposing it to a client. These are the projects in operation:
How we work
with an organization
- Process diagnosis: what's done today, how long it takes, where time is lost, and whether it's measurable. If it isn't, the first deliverable is the measurement.
- Scoped use case: just one, with a success criterion defined before starting.
- Measurable pilot: against the process's baseline, not against an expectation.
- Production and operation: monitoring, knowledge base maintenance, and continuous adjustment.
Where we work
Coresis operates from Bogotá and works remotely with organisations across Colombia: Medellín, Cali, Barranquilla, Bucaramanga and Cartagena among others. We do not keep offices in every city and do not need to: an AI agent project is defined in working sessions and delivered on the client's systems, not at a shared desk.
What does matter about location is the time zone and the legal framework. A shared working day means an incident can be handled the same day, and being incorporated in Colombia means the contract and the data processing fall under Colombian law. For a public institution that is not a detail: it is a requirement.
Frequently asked questions
What does an artificial intelligence agency actually do?
It should do three things: identify which processes justify an AI solution and which don't, build and integrate the solution with the systems you already operate, and leave it measured and maintainable. If a proposal starts with the technology and not the process, it's worth reviewing.
- That it starts from the process, not the tool.
- That it's explicit about where the data stays and with which model.
- That it separates the three costs: build, consumption, and operation.
- That it defines how the result will be measured before building.
- That it can show systems sustained over time, not just demos.
- That it hands operation to the in-house team when the client wants that.
Is Coresis an AI agency or an engineering company?
It's a software engineering team that works with AI as one more capability in the stack. The practical difference is the starting point: we arrive from integration with production systems, not from building prototypes.
Can you guarantee that data won't leave our infrastructure?
It depends on the design. Orchestration with n8n can be self-hosted and open models can be deployed on your own infrastructure, which keeps sensitive data from leaving the perimeter. With commercial models via API that doesn't apply, and it has to be decided case by case.
How much does it cost to work with an AI agency in Colombia?
The cost splits into build (one-time, depends on how many systems need to be integrated), consumption per model interaction (variable), and operation (recurring, usually underestimated). We deliver the phased estimate after the initial diagnosis.
Do you work with public entities?
Yes. Coresis S.A.S. is incorporated in Colombia and has delivered platforms for national and district entities. We can take part in public procurement processes and adapt to the entity's oversight and delivery requirements.
Where does a project start?
With a diagnosis of the process, not the model. If the process can't be measured today, the first useful deliverable is the measurement. From there we define a scoped use case, test it, and only then move it to production.
Should we evaluate an AI use case for your organization?
Tell us what process you want to solve and we'll review whether it justifies an AI solution, or if it's better solved another way.